add gpu info
Browse files- backend-cli.py +2 -1
- src/display/utils.py +3 -5
backend-cli.py
CHANGED
@@ -440,7 +440,8 @@ if __name__ == "__main__":
|
|
440 |
"mistralai/Mixtral-8x22B-v0.1", "mistralai/Mixtral-8x22B-Instruct-v0.1", "alpindale/WizardLM-2-8x22B",
|
441 |
"CohereForAI/c4ai-command-r-plus"] # Use model from arguments
|
442 |
debug_task_name = ['mmlu', 'selfcheckgpt'] # Use task from arguments
|
443 |
-
precisions = ['4bit', '
|
|
|
444 |
task_lst = TASKS_HARNESS.copy()
|
445 |
for precision in precisions:
|
446 |
for debug_model_name in debug_model_names:
|
|
|
440 |
"mistralai/Mixtral-8x22B-v0.1", "mistralai/Mixtral-8x22B-Instruct-v0.1", "alpindale/WizardLM-2-8x22B",
|
441 |
"CohereForAI/c4ai-command-r-plus"] # Use model from arguments
|
442 |
debug_task_name = ['mmlu', 'selfcheckgpt'] # Use task from arguments
|
443 |
+
# precisions = ['4bit', '8bit']
|
444 |
+
precisions = ['float32', 'float16']
|
445 |
task_lst = TASKS_HARNESS.copy()
|
446 |
for precision in precisions:
|
447 |
for debug_model_name in debug_model_names:
|
src/display/utils.py
CHANGED
@@ -105,16 +105,14 @@ for task in Tasks:
|
|
105 |
# System performance metrics
|
106 |
auto_eval_column_dict.append([f"{task.name}_end_to_end_time", ColumnContent, ColumnContent(f"{task.value.col_name} {E2Es}", "number", True)])
|
107 |
auto_eval_column_dict.append([f"{task.name}_batch_size", ColumnContent, ColumnContent(f"{task.value.col_name} {BATCH_SIZE}", "number", True)])
|
|
|
108 |
auto_eval_column_dict.append([f"{task.name}_gpu", ColumnContent, ColumnContent(f"{task.value.col_name} {GPU_Name}", "str", True)])
|
|
|
109 |
if task.value.benchmark in MULTIPLE_CHOICEs:
|
110 |
continue
|
111 |
-
auto_eval_column_dict.append([f"{task.name}_prefilling_time", ColumnContent, ColumnContent(f"{task.value.col_name} {PREs}", "number", False)])
|
112 |
auto_eval_column_dict.append([f"{task.name}_decoding_throughput", ColumnContent, ColumnContent(f"{task.value.col_name} {TS}", "number", True)])
|
113 |
|
114 |
-
auto_eval_column_dict.append([f"{task.name}_gpu_mem", ColumnContent, ColumnContent(f"{task.value.col_name} {GPU_Mem}", "number", True)])
|
115 |
-
auto_eval_column_dict.append([f"{task.name}_gpu_power", ColumnContent, ColumnContent(f"{task.value.col_name} {GPU_Power}", "number", False)])
|
116 |
-
auto_eval_column_dict.append([f"{task.name}_gpu_temp", ColumnContent, ColumnContent(f"{task.value.col_name} {GPU_TEMP}", "number", False)])
|
117 |
-
auto_eval_column_dict.append([f"{task.name}_gpu_util", ColumnContent, ColumnContent(f"{task.value.col_name} {GPU_Util}", "number", True)])
|
118 |
|
119 |
# Model information
|
120 |
auto_eval_column_dict.append(["model_type", ColumnContent, ColumnContent("Type", "str", False)])
|
|
|
105 |
# System performance metrics
|
106 |
auto_eval_column_dict.append([f"{task.name}_end_to_end_time", ColumnContent, ColumnContent(f"{task.value.col_name} {E2Es}", "number", True)])
|
107 |
auto_eval_column_dict.append([f"{task.name}_batch_size", ColumnContent, ColumnContent(f"{task.value.col_name} {BATCH_SIZE}", "number", True)])
|
108 |
+
auto_eval_column_dict.append([f"{task.name}_gpu_mem", ColumnContent, ColumnContent(f"{task.value.col_name} {GPU_Mem}", "number", True)])
|
109 |
auto_eval_column_dict.append([f"{task.name}_gpu", ColumnContent, ColumnContent(f"{task.value.col_name} {GPU_Name}", "str", True)])
|
110 |
+
auto_eval_column_dict.append([f"{task.name}_gpu_util", ColumnContent, ColumnContent(f"{task.value.col_name} {GPU_Util}", "number", True)])
|
111 |
if task.value.benchmark in MULTIPLE_CHOICEs:
|
112 |
continue
|
113 |
+
# auto_eval_column_dict.append([f"{task.name}_prefilling_time", ColumnContent, ColumnContent(f"{task.value.col_name} {PREs}", "number", False)])
|
114 |
auto_eval_column_dict.append([f"{task.name}_decoding_throughput", ColumnContent, ColumnContent(f"{task.value.col_name} {TS}", "number", True)])
|
115 |
|
|
|
|
|
|
|
|
|
116 |
|
117 |
# Model information
|
118 |
auto_eval_column_dict.append(["model_type", ColumnContent, ColumnContent("Type", "str", False)])
|